[forgotpwd16]

There's also the issue, which is also noted by the author, that LLM-optimization quite often becomes, when shouldn't be just that, token-minimization.